Re: Xalan-J XSL 3 test cases submission

Hi John,
   Thanks for the thoughts.

As per your advice, I'll try to test Xalan-J's XSLT3 implementation
with respect to W3C XSLT3 tests. I shall share my findings to this
list, when I'm ready with those details.

On Tue, Apr 1, 2025 at 10:51 PM John Lumley <john@saxonica.com> wrote:
>
> Without revealing what the tests are and the anticipated results, I’m afraid this is of no use whatsoever in helping potential developers/users decide how mature your XSLT processor is or even whether it seems to be going in the right direction. Perhaps they indicate you’re making progress, but I can assure you it’s very easy to get the test cases wrong, especially in the all-important edge conditions.
>
> And there is a throughly developed and publicly available set of some O(10,000) XSLT tests available in the W3C test set, as well as the O(20,000) XPath tests. Results against those tests will be much more meaningful, and indeed developing an XSLT compiler incrementally against selected sections of those tests is a very effective design methodology.
>
> No-one expects you to pass all tests at first, or even twentieth, go, but for example passing most of the xsl:choose tests is a good indicator that you’ve nailed that (fairly simple) section. Indeed when developing the XSLT compiler for SaxonJS, I started by building the code for xsl:choose compilation and testing against the extant test set, and then built outwards over an already tested ‘core’.
>
> I would seriously recommend you move to running portions of the W3C XSLT test suite, and then reporting those results.
>
> John Lumley
>
> Sent from my iPad


-- 
Regards,
Mukul Gandhi

Received on Wednesday, 2 April 2025 11:01:59 UTC