Re: [xsl] Request for XSLT2.0 / XML Schema triples

[Soren Kuula <dongfang@xxxxxxxxxxx>]

Michael Kay wrote:

> I'll try and dig out a few examples for you. Most of my test cases have
> either an input schema or an output schema but not both, and others are
> proprietary to clients, so I don't actually have all that much material.
>
> Do you assume there are no templates applied to temporary trees? One of the
> difficulties in doing this kind of analysis is that a template rule with
> match="E" is not necessarily matching an element that conforms to the
> definition of E in the input schema. Similarly a template that constructs an
> element E might be writing it to a temporary tree rather than to the final
> output tree. Of course if you are only after an approximate verdict you can
> afford the luxury of making guesses on such things. 
>  
Hi,

Thanks for the effort.

I'm being conservative, that is, if it says valid, then it must be 
valid, and I'm in trouble when someone's nuke powerplant blows up 
because it wasn't.
My escape path is then, of course, just to say invalid if in doubt.

(If I had the liberty to approximate in both directions, I'd end up with 
an algorithm that couldn't quite be trusted when saying yes, and ditto 
when saying no. Absolutely useless, that is.)

And I didn't even have the time yet to look into temporary trees (I 
might have to make to with sketchy arguments on those). Reference and 
uniqueness constraints are way too nasty to analyze statically too, so 
the algorithms's verdict is really with respect to an output schema 
where that kind of stuff is removed.

So -- it is fine, just fine, with single input single output examples.

Soren