W3C home > Mailing lists > Public > xproc-dev@w3.org > November 2016

Re: Added support for Saxon 9.7 to Calabash

From: Florent Georges <fgeorges@fgeorges.org>
Date: Sun, 6 Nov 2016 16:10:34 +0100
Message-ID: <CADyR_r29rhrxj2TR8HKQB=w+r5HT-dfvF9A0eSJ=Uj4cxo3Mrg@mail.gmail.com>
To: XProc Dev <xproc-dev@w3.org>

A quick update on this.  To summarize, I created a branch to support
Saxon 9.7: https://github.com/fgeorges/xmlcalabash1/commits/saxon97,
where there were still the following tests failing:

1/ Schematron validation fails (testValidsch001 to testValidsch005)
2/ Default collection not set (testXquery006 and testXslt2003)
3/ Base URI on secondary documents (testXslt006 and testXslt2002)

On 1/, I was puzzled with the fact Schematron validation was requiring
SaxonEE.  Actually, the test used for detecting whether "schema-
awareness" was required (in ValidateWithSCH.java, line 96) is not
correct anymore.  In Saxon 97, the method getSchemaType() does not
return null anymore, but "xs:untyped".  The change I made is
compatible with both version (if null, or xs:untyped, then it is not
SA): http://github.com/fgeorges/xmlcalabash1/commit/d8c9dbcb3da7782.

So these 5 tests pass now.

On 2/, I could actually not make the simple query "collection()" work
on Saxon HE 9.7, using s9api.  I have reported this to the Saxon list:

Therefore, on the above branch, the following tests are the only tests
to fail, still:

- testXquery006
- testXslt2003
- testXslt006
- testXslt2002


Florent Georges

On 29 October 2016 at 19:59, Florent Georges wrote:
> Hi,
> I have a project depending on both Saxon and Calabash.  I have upgraded it to
> Saxon 9.7, but I am blocked by the lack of support of Calabash for Saxon 9.7.
> So I created a new branch and adapted Calabash codebase to Saxon 9.7.  You can
> find the changes at https://github.com/fgeorges/xmlcalabash1/commits/saxon97.
> The test suite reports: "717 tests completed, 9 failed"  Out of these 9, 5 are
> because I use HE and they require EE (because it is about Schematron validation,
> I doubt upgrading Saxon would break them, but I can't test them):
> - Optional.testValidsch001
> - Optional.testValidsch002
> - Optional.testValidsch003
> - Optional.testValidsch004
> - Optional.testValidsch005
> 2 tests are failing because the default collection is not set:
> - Optional.testXquery006
> - Optional.testXslt2003
> And finally, 2 tests are failing because the base URI is not set correctly on a
> secondary document created by xsl:result-document in a pipeline:
> - Required.testXslt006
> - Optional.testXslt2002
> This last error is due to Saxon setting the wrong system ID in that case.  Given
> what's in ElementCreator.java at line 350, it looks like it sets the document
> node system ID to the one of its root element (inheriting it from the stylesheet
> source in the pipeline file), instead of using the one from the `href` attribute
> on xsl:result-document.  Not sure this is because of Saxon or of a
> change in 9.7.
> The following is the stack trace from the point where the wrong system
> ID is set:
> at net.sf.saxon.event.ComplexContentOutputter.setSystemId(ComplexContentOutputter.java:134)
> at net.sf.saxon.event.ProxyReceiver.setSystemId(ProxyReceiver.java:40)
> at net.sf.saxon.event.ComplexContentOutputter.setSystemId(ComplexContentOutputter.java:134)
> at net.sf.saxon.expr.instruct.ElementCreator.processLeavingTail(ElementCreator.java:350)
> at net.sf.saxon.expr.instruct.ElementCreator.processLeavingTail(ElementCreator.java:311)
> at net.sf.saxon.expr.instruct.Instruction.process(Instruction.java:149)
> at net.sf.saxon.expr.instruct.ResultDocument.processInstruction(ResultDocument.java:470)
> at net.sf.saxon.Configuration.processResultDocument(Configuration.java:2132)
> at net.sf.saxon.expr.instruct.ResultDocument.process(ResultDocument.java:382)
> So it is definitely not 100% complete, but except for these 2 corner cases (WRT
> the default collection and secondary outputs in XSLT), it seems to be working
> fine.  It is maybe time to add beta support for 9.7, released 1 year ago and to
> which most projects relying on Saxon seem to have been moved to, even if only
> 99% of the test suite pass.  Having this support in beta status for one or a few
> versions of Calabash would help gather feedback as well.
> Not sure whether it is possible to create a pull request for a new branch, but
> happy to investigate if you want me to.
> Regards,
> --
> Florent Georges
> http://fgeorges.org/
> http://h2oconsulting.be/
Received on Sunday, 6 November 2016 15:11:28 UTC

This archive was generated by hypermail 2.3.1 : Sunday, 6 November 2016 15:11:29 UTC